Spectral correction study to reduce the influence of sample surface morphology on laser-induced breakdown spectroscopy

Abstract

For in situ detection via laser-induced breakdown spectroscopy (LIBS), the effect of the sample surface morphology matrix exacerbates spectral variations, leading to inaccurate and unstable analysis results, even under ideal laboratory conditions. Thus, it is essential to consider the influence of sample surface morphology on LIBS to obtain optimal detection results. Therefore, to reduce this influence, herein, we proposed spectral correction methods to correct LIBS spectra based on our previous research results on the relationship between LIBS characteristic spectral intensity and physical parameter of the sample surface morphology, where the parameter was the angle θ between the tangent of the sample surface and the horizontal direction. The results showed that the fluctuations in the corrected spectra were reduced and the relative standard deviation (RSD) was less than 10%, and the spectral intensities were close to that of the samples with a flat surface morphology. To assess the effectiveness of the proposed methods in improving the quantitative analysis accuracy, Cr, Ni, and Mn elements in standard stainless-steel samples were analyzed by partial least squares regression (PLSR). We established the model using the spectra of standard samples with a flat surface morphology as the training set, and the results showed that the accuracy was very low when the test set was the spectra of samples without a flat surface morphology, with R² of 0.781, 0.779, and 0.695 for Cr, Ni, and Mn elements, respectively. In contrast, when the test set was the corrected spectra, R² increased to 0.950, 0.925, and 0.825 for Cr, Ni, and Mn elements, respectively, and the RMSE and MRE were reduced. When the corrected spectra of standard samples without a flat surface morphology were added to the training set, the accuracy was improved, with R² of 0.978, 0.939, and 0.929 for the three elements and their RMSE and MRE were minimum.